Automated Unimpaired Hydrologic Metric Scaling for California Streams

Transcription

1 Automated Unimpaired Hydrologic Metric Scaling for California Streams Abstract The following report is the beginning of a larger study to provide unimpaired daily hydrographs for streams in the state of California. The larger project objective is to create an ArcGIS based tool that will automatically produce streamflow estimates for any reach in California. This future method will use reference gauges and to-be-determined scaling methods. This paper reviews the available scaling methods and compares them for accuracy and dependability. Data was obtained from ArcGIS Online and Dr. Belize Lane at Utah State University. ArcGIS was used to organize, calculate, and select data. A large portion of the research involved coding scenarios in Visual Basic and Excel to test different scalar methods. Results of the project are preliminary but show that traditional scaling approaches can provide accurate predictions with some limitations. They also show that a classification system for providing metrics may be better at predicting hydrograph shapes than traditional methods alone. By Karl Christensen GIS Water Resources CEE 6440 December 2017

2 Introduction The objective of this research project is to provide unimpaired flow metrics for any stream reach in California. These reference flow metrics are characteristics of the timing, magnitude, duration, frequency and rate of change (Poff et al, 1997) of daily streamflow time series in the absence of major alterations by dams, diversions, and land use changes. Reference flow metrics have been linked to ecological integrity and can be used to guide flow management decisions to restore or retain ecological objectives. Predicting reference flow metrics in stream reaches without local gauge data is an existing challenge. Here, several alternative approaches have been evaluated for scaling streamflow data from existing reference gauges to ungauged locations and their ability to predict a set of reference flow metrics is compared. An unimpaired daily flow regime directly affects stream ecology. Hydrology and aquatic biodiversity have been linked via four key mechanisms: a) flow is a major determinant of the habitat, a key driver of the aquatic composition, b) aquatic species have evolved life-history strategies in response to the natural flow regime, c) the natural pattern of the longitudinal and lateral connectivity in the river system is important for supporting populations of aquatic species and d) the invasion and success of non-native species is facilitated by alterations to streamflow (Bunn and Arthington, 2002). Alteration of the natural flow regime often leads to ecological degradation and the shifting of species assemblage away from native species (Chinnayakanahalli, 2010). Several traditional scaling methods are established and used to predict streamflow when a reference stream gauge is not available. Farmer and Vogel (Farmer and Vogel, 2012) list three of these methods: 1. Scaling flows by the Drainage-Area Ratio (DAR) technique 2. Scaling flows by average streamflow 3. Scaling flow by average and standard deviation of streamflow The DAR scalar is a common scaling method because it only requires catchment areas of the reference and prediction sites and streamflow data from the reference site. Methods 2 and 3 require streamflow data estimates from both the reference and prediction sites but no drainage areas. While these methods are useful for direct interpolation, they do not account hydrologic variability in the streams being compared. Potential variability not accounted for in streamflow averages or catchment areas may negatively influence the results. Stream flow variability can be better predicted when streams are compared across an entire region. Hersch and Maidment (2007) created a classification scheme that distinguished geographic regions of streams with similar attributes related to water quality, climatology, hydrology & hydraulics, geomorphology & physical processes, and biology. Five hydrologic regions were identified for the state of Texas: North-Central Texas, West Texas, East Texas, 2

3 Lower Rio Grande Basin, and South-Central Texas (see Figure 1 below). Their results allowed further hydrologic analysis to recognize which streams will have similar behavior. Figure 1 - Texas Classifications: Five Regions (Maidment and Hersch, 2007) Similar to Maidment and Hersch, Lane et al (2017) distinguished nine hydrologic classes for the state of California comparing 20 different attributes. Unlike Maidment and Hersch, Lane et al did not attempt to build geographic boundaries but instead depended upon a statistical analysis of each individual stream s attributes (see Figure 2). The resulting stream classes can be seen in Figure 3 and show a much messier regionalization with a more heterogeneous distribution. Each stream class is named according to the driving hydrologic conditions: Class 1 - Snowmelt, Class 2 - Low-volume snowmelt and rain, Class 3 - High-volume snowmelt and rain, Class 4 - Rain and seasonal groundwater, Class 5 - Winter storms, Class 6 - Groundwater, Class 7 - Perennial groundwater and rain, Class 8 - Flashy, ephemeral rain, and Class 9 - High elevation low precipitation. Class 9 is not shown in Figure 3 because it was developed after the original publication of Lane et al Figure 2 Hydrologic Indices (Lane et. al, 2017) 3

4 Lane et al further defined the classes by creating Dimensionless Reference Hydrographs (DRH) for each stream and an average DRH for each class. The DRH values can be interpreted as metrics such as timing of yearly peak flows, duration of peak flows, lowest mean annual flows, and other hydrologic characteristics. As part of an assigned class, each DRH also holds certain physical and climatic catchment controls that allow most streams in California to be classified. Streams of the same classification can be assumed to hold similar DRH patterns and provide a foundation for developing alternative scaling methods. Figure 3 - California Stream Classes (Lane et. al, 2017) Method of Work Data from multiple sources were used to test each of the traditional and classification based scaling methods. Data provided by Dr. Belize Lane at Utah State University included the DRH values for individual gauge stations and class averages, actual stream flow data for gauges, and monthly & annual flow estimates for each predicted gauge. Drainage areas were calculated in ArcGIS Pro using the National Elevation Dataset at 30m resolution and the ArcGIS hydrology tool package. All data was combined first in ArcGIS Pro and then tabulated in Excel. The automation of calculating each method was accomplished using Visual Basic (VBA). The traditional and classification scaling methods were expanded into 10 separate scenarios shown below in Table 1. Each scenario was added into the VBA code to predict a 20-year daily streamflow time series. The classification methods utilized DRH Values and Annual/Monthly Averages from the USGS (via Dr. Lane); the traditional methods utilized actual daily flows and calculated drainage areas depending on the scenario. 4

5 Table 1 - Scenarios Methods (Classification Method - Blue, Traditional Method - Green, Actual Daily Time Series - Red) TYPE SCENARIO TIME SERIES SCALARS CLASSIFICATION 1 Aggregate DRH Values Annual Averages CLASSIFICATION 2 Aggregate DRH Values Monthly Averages CLASSIFICATION 3 Nearest 1 DRH Values Annual Averages CLASSIFICATION 4 Nearest 1 DRH Values Monthly Averages CLASSIFICATION 5 Nearest 3 DRH Values Annual Averages CLASSIFICATION 6 Nearest 3 DRH Values Monthly Averages TRADITIONAL 7 Nearest 1 Daily Flows Drainage Area Ratio TRADITIONAL 8 Nearest 1 Daily Flows Annual Average Ratio TRADITIONAL 9 Nearest 1 Daily Flows Monthly Average Ratio TRADITIONAL 10 Nearest 1 Daily Flows Standard Deviation Ratio NO SCALING Actual Prediction Site Daily Flows N/A Prediction and reference gauge sites were chosen in ArcGIS Pro by selecting an area with at least four gauge sites of the same class. Centermost sites were used as prediction sites and gauges nearby as reference sites (see Figure 4). USGS Gauge identification numbers, such as those shown in Figure 4, were then used as inputs in the Excel input worksheet. The VBA main code then ran all 10 scenarios for the 4 gauge sites and produced a 20-year daily time series for each scenario. A copy of the VBA Code can be found in Appendix A. Class 1 Gauges near Yosemite National Park, CA Class 1 Reference Gauges (Sites to be referenced) Class 1 Prediction Gauge (Site to be predicted) Figure 4 - Class 1 Gauges near Yosemite National Park 5

6 The hand calculations for building each time series and scalar can be seen in Figures 5 & 6. Notice that the time series are label A through D and the scalars are labeled 1 through 6. The scenarios then have labels such as A-1, B-2, D-4, and so on, depending on the combination of series and scalars. These calculations were used to confirm the accuracy of the program and spreadsheet values. Figure 5 - Hand Calculations Part 1 6

7 Figure 6 - Hand Calculations Part 2 Results Results from each scenario were compared to the actual daily flows of the prediction site. Traditional scaling methods worked very well when reference gauges within the same class were chosen. Figure 7 on the following page shows all 10 scenario results for the 1969 water-year at USGS gauge Note how the runoff peaks estimated by the DRH values were of a similar duration to the actual runoff, but shifted to earlier in the year (Scenarios 1 through 6). This shift is likely due to seasonal shifts made over the 20-year period used to define the DRH values. Traditional methods were found to be very accurate at predicting the daily flows for the entire year. Such a high level of accuracy was assumed to correlate directly with the fact that all reference gauges were both near the prediction site, and of a similar class. 7

8 Figure 7 - USGS Gauge Results 8

9 Limitations for the traditional scaling methods were found when predicting streamflow between gauges from different classes. For example, a stream that is primarily fed by snowmelt is shown by USGS Gauge It has a hydrograph with one large peak from snowmelt runoff in the spring and relatively low flow in the late summer and fall (see Figure 8). Figure 8 - Daily Streamflow Gauge Gauge shows a Class 3 stream fed by snowmelt and seasonal rainfall that has a runoff peak along with variable seasonal flows (see Figure 9). Figure 9 - Daily Streamflow Gauge When the DAR method is used to estimate one year of daily flows for Gauge by using Gauge the shape of the hydrograph is incorrect (see Figure 10). Figure 10 - Daily and Predicted Streamflow Gauge

10 Conclusion The results discussed thus far show an indeterminate conclusion without further data. The DRH values did not produce more accurate results than the traditional methods but the traditional methods were only supremely accurate when used within a class. The classification metrics are more consistent at producing correctly shaped hydrographs over a series of years but lack accuracy for specific days. ArcGIS proved to be an invaluable resource in organizing the gauge data points and selecting the reference and prediction sites. While the original objective to create flow metrics for every stream in California was not met, the results provided will provide a strong basis and the tools necessary to create those metrics in the future. Direction for Future Work The ultimate goal of this project is to create an ArcGIS tool in python to calculate daily flows of ungauged sites. While this goal will be attainable in future efforts, the initial analysis of scaling methods were too time intensive to be completed within a single semester. Further work on this project will involve comparing additional results and choosing an optimal method or combination of methods to create satisfactory predictions. A fully debugged and user-friendly tool will likely not be completed until spring of next year. If successful, this tool could be used to predict flows across the state of California and serve as an example for other project areas. 10

11 REFERENCES Chinnayakanahalli, K. J. (2010). Characterizing Ecologically Relevant Variations in Streamflow Regimes. All Graduate Theses and Dissertations, Paper 561 Farmer, W. H., and Vogel, R. M. (2012). Performance-Weighted Methods for Estimating Monthly Streamflow at Ungauged Sites. Department of Civil and Environmental Engineering, 200 College Avenue, Medford, MA 02155, United States Hersch, E. S., and Maidment, D. R. (2017). An Integrated Stream Classification System for Texas. CENTER FOR RESEARCH IN WATER RESOURCES Bureau of Engineering Research The University of Texas at Austin J.J. Pickle Research Campus Austin, TX , < (Oct. 1, 2017) Lane, B. A., Dahlke, H. E., Pasternack, G. B., and Sandoval-Solis, S. (n.d.). Revealing the diversity of Natural Hydrologic Regimes in California with Relevance for Environmental Flows Applications. Journal of the American Water Resources, 1 20 Poff, N.L., J.D. Allan, M.B. Bain, J.R. Karr, K.L. Prestegaard, B.D. Richter, R.E. Sparks, and J.C. Stromberg, The Natural Flow Regime. BioScience 47(11): , DOI: /

12 Visual Basic Code used to run each scenario in Excel APPENDIX A 12

13 13

14 14

15 15

16 16

17 17

18 18

19 19

20 20

21 21

22 22

23 23